Segmentation Based Multilevel Wide Band Compression for SAR Images Using Coiflet Wavelet

P.Subashini & M.Krishnaveni
International Journal of Image Processing (IJIP), Volume (5) : Issue (2) : 2011 177
Segmentation Based Multilevel Wide Band Compression for SAR
Images Using Coiflet Wavelet
Dr.P.Subashini mail.p.subashini@gmail.com
Associate Professor
Department of Computer Science
Avinashilingam Deemed University for Women
Coimbatore, 641 043, India
M.Krishnaveni krishnaveni.rd@gmail.com
Research Assistant
Department of Computer Science
Avinashilingam Deemed University for Women
Coimbatore, 641 043, India
Abstract
Synthetic aperture radar (SAR) data represents a significant resource of information for a large
variety of researchers. Thus, there is a strong interest in developing data encoding and decoding
algorithms which can obtain higher compression ratios while keeping image quality to an
acceptable level. In this work, results of different wavelet-based image compression and
segmentation based wavelet image compression are assessed through controlled experiments
on synthetic SAR images. The effects of dissimilar wavelet functions, number of decompositions
are examined in order to find optimal family for SAR images. The choice of optimal wavelets in
segmentation based wavelet image compression is coiflet for low frequency and high frequency
component. The results presented here is a good reference for SAR application developers to
choose the wavelet families and also it concludes that wavelets transform is rapid, robust and
reliable tool for SAR image compression. Numerical results confirm the potency of this approach
Keywords: Image Processing, Compression, Segmentation, Wavelets, Quality Measures, SAR
Images
1. INTRODUCTION
The necessity of efficient and consistent compression techniques for remote sensing imagery is
increasing rapidly on the bases of network transmission [2]. In image processing domain the
neglection of peculiar characteristics of the data accounts a lot in poor recognition system [9]. The
usual approach to deal with this problem is to resort to edge-preserving filters. Segmentation,
compression and filtering should be carried out jointly for extraction and better representation of
most relevant features. Therefore a fundamental approach to image filtering and compression
requires the prior segmentation of the image in homogeneous regions [5].This work includes the
strong background of research result proposed by Zhaohui Zeng and Ian Cumming for SAR
image compression using DWT [15]. Here wavelet-based despeckling is used implicitly to
distinguish between noise components (high and low) and its state boundaries[4][2]. This work
aims at studying and quantifying the potential advantages provided by wavelet towards
segmentation based image compression. This work is investigated from few SAR image
collections like ERS Synthetic Aperture Radar (SAR) imagery , ESA,JERS and radarsat . The
paper follows as. Section 2 explains the compression schemes needed for SAR segmentation.
Section 3 deals with wavelet analysis and its metrics which comprises the performance measures
of wavelet families. Section 4 describes the compression coding schemes based on
segmentation. Section 5 presents and discusses the results of a number of experiments, and
Section 6 draws conclusions with possible scenario.

2. SAR IMAGE WAVELET COMPRESSION
In the scientific literature concerning SAR system, the term compression is often used to indicate
the data processing which allows focusing the received echoes. SAR domain compression[3][14]
is carried out off-line at the base station, which can leverage on more time and computation
power. So image compression became an operating instrument, for scientific research[9]. In basic
literature evaluation is done on quality metrics, neglecting the visual effects and the
consequences on the image interpretability. In order to attain above conditions and limitations of
SAR, new compression techniques on wavelet transform is characterized with good time-
frequency localization, multiresolution representation low-complexity implementation through filter
banks [5]. Since the wavelet transform is a bandpass filter with a known response function (the
wavelet function), it is possible to reconstruct the original time series using either deconvolution or
the inverse filter [15]. This is straightforward for the orthogonal wavelet transform (which has an
orthogonal basis), but for the continuous wavelet transform it is complicated by the redundancy in
time and scale. However, this redundancy also makes it possible to reconstruct the time series
using a completely different wavelet function, the easiest of which is a delta (δ) function[1]. In this
case, the reconstructed time series is just the sum of the real part of the wavelet transform over
all scales which is stated in eqn(1)
∑=
ℜ
=
J
j j
jntj
n
s
sW
OC
x
0
2/1
0
2/1
)}({
)(ψ
δδ
δ
----(1)
The factor )0(=nψ removes the energy scaling, while the sj
1/2
converts the wavelet transform to
an energy density. The factor Cδ comes from the reconstruction of a δ function from its wavelet
transform using the function )(nn =ψ . This Cδ is a constant for each wavelet function .Note that
if the original time series was complex, then the sum of the complex Wn(s) would be used instead
[3][5]. To derive Cδ for a new wavelet function, first assume a time series with a δ function at time
n = 0, given by xn = δn. This time series has a Fourier transform sωk = N-1, constant for all k.
Substituting sωk into (1), at time n = 0 (the peak), the wavelet transform becomes
)(*
1
)(
1
0
^
k
N
k
s
N
SW ωψδ ∑
−
=
= ---------------------- (2)
The reconstruction eqn (2) then gives eqn (1) .The Cδ in eqn (1) is scale independent and is a
constant for each wavelet function.
∑ ∑
−
= =
=
1
0 0
2
2
)(N
x
J
j j
jntj
s
sW
NC δ
δδ
σ -------(3)
The total energy is conserved under the wavelet transform, and the equivalent of Parseval’s
theorem for wavelet analysis is given in eqn(3).Both eqn (1) and eqn (3) should be used to check
wavelet routines for accuracy and to ensure that sufficiently small values of sj and δj have been
chosen.
3. WAVELET ANALYSIS
This section describes the method of wavelet analysis, includes a discussion of different wavelet
functions, and gives details for the analysis of the wavelet power spectrum. Results in this section
are adapted to Segmentation based wavelet compression from the continuous formulas given in
coiflets. The wavelet bases performed over SAR images are as follows. Haar wavelet is the
simplest of the wavelet transforms[10]. This transform cross-multiplies a function against the Haar
wavelet with various shifts and stretches, like the Fourier transform cross-multiplies a function
against a sine wave with two phases and many stretches. Daubechies wavelets are a family of
orthogonal wavelets defining a discrete wavelet transform and characterized by a maximal
number of vanishing moments for some given support. Meyer's wavelet construction is

fundamentally a solvent method for solving the two-scale equation. Symlet wavelet is only nearly
symmetric, and is not exactly symmetrical. Coiflets are discrete wavelets designed by Ingrid
Daubechies, to have scaling functions with vanishing moments. Biorthogonal wavelet is a
wavelet where the associated wavelet transform is invertible but not necessarily orthogonal.
Reverse biorthogonal is a spline wavelet filters.
To verify the validity of the wavelet families the results are compared based on PSNR ratio and
MSE. With the extension of the work, subjective evaluation is also done based on the
decomposition level. Compression is been the main phenomena for the evaluation job [8][12].
The DWT was used with coiflets, least asymmetric compactly-supported wavelet with eight
vanishing moments with four scales. The 120 x 120 pixel region SAR images are used for
applying wavelet families. They were contaminated with speckle noise[2][3].
Some of the parameters taken for analysis of wavelet on SAR images are Mean Square Error,
Peak Signal to Noise Ratio, Compression Score and Recovery Performance.
FIGURE 1: (a)Haar wavelet (b) Daubechies wavelet (c) Meyer Wavelet (d) Symlet wavelet
(e) Coiflets wavelet (f) Bi orthogonal wavelet (g) Reverse Bi orthogonal

FIGURE 2: (a) Peak Signal to Noise Ratio (b) Mean Square Error Rate (c) Recovery Performance (d)
Compression Score
A sample of one SAR image is subjectively explained by the results in the above figure (1) and
objectively in figure (2). From the above got results the coiflets of wavelet families out performs
the rest of the wavelet families.
4. SEGMENTATION BASED WAVELET COMPRESSION
The idea behind wavelet image compression, like that of other transform compression techniques
is, it removes some of the coefficient data from the transformed image [8]. Encoding may be
applied to the remaining coefficients. The compressed image is reconstructed by decoding the
coefficients, if necessary, and applying the inverse transform to the result[17]. No much image
information is lost in the process of removing transform coefficient data.
FIGURE 3: Segmentation based encoding schemes
This process of image compression through transformation and information removal provides a
useful framework for understanding wavelet analysis[8]. Mathematical references develop
Denoise Compress
Segment Denoise
Fixed/adapt
Compress
Fixed/adapt
M+T+N M+T
M+T+N T+N
Code Map
T

multiresolution analysis through the definition of continuous scaling and wavelet functions[18]. In
order to compare the performance of wavelet and contourlet transform in our proposed
compression scheme, we compute the compression ratios (CR) for various quantization levels.
CR is defined as the ratio between the uncompressed size and the compressed size of an image
[13]. To compute the CR in a fairly way, the original image is encoded using the resulting number
of bits it saved. This compression eliminates almost half the coefficients, yet no detectable
deterioration of the image appears. The compression procedure contains three steps: (i)
Decomposition. (ii) Detail coefficient thresholding[16] ,for each level from 1 to N, a threshold is
selected and hard thresholding is applied to the detail coefficients, and(iii) Reconstruction.
5. PERFORMANCE EVALUATION
Fig 4 and 5 shows PSNR and compression ratio curves for various SAR images[3] through
compression techniques. Fig 6 and 7 shows the Normalized cross correlation and Normal
Absolute error rate for the sample SAR images between two methods wavelet based
compression (WBC) and segmentation based wavelet compression (SBWC). The image size
held is 256x256. These curves plot compression ratio versus image quality, as represented by
PSNR. The most desirable state on this graph is the upper right quadrant, where high
compression ratios and good image quality live.
From the Figure 6 and 7 the Cross-correlations help identify variables which are leading
indicators of other variables or how much one variable is predicted to change in relation the other
variable. The identified method is justified in all the above taken parameters which objectively
states the phenomena in prominent manner
1 2 3 4 5 6 7 8 9
18
20
22
24
26
28
30
32
34
36
38
Number of Images (256 x256 pixels)
PSNR(db)
Comparison of PSNR values for two methods
wbc
sbwc
FIGURE 4: Comparison of PSNR for WBC and SBWC

1 2 3 4 5 6 7 8
65
70
75
80
85
90
95
Number of images (256x256 pixels)
CompressionRatioin%
Comparison of Compression ration for two methods
WBC
SBWC
FIGURE 5: Comparison of CR for WBC and SBWC
1 2 3 4 5 6 7 8
0.9
0.95
1
1.05
1.1
1.15
1.2
1.25
Number of images (256x256)
Normalizedcrosscorrelation
Comparison of NCC for two methods
SBWC
WBC
FIGURE 6: Comparison of NCC for WBC and SBWC

1 2 3 4 5 6 7 8 9
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
Number of images (256x256)
NormalAbsoluteError
Comparison of NAE for two methods
SBWC
WBC
FIGURE 7: Comparison of NAE for WBC and SBWC
6. CONCLUSION
The objective specified is clearly evaluated by segmentation based image compression
performance, in accounting compression ratio, PSNR and time needs. It is oriented toward the
operational naval application intelligent imagery system. The images were evaluated
quantitatively and qualitive assessments are done by metrics in order to establish the impact of
wavelet and the need of segmentation based approach for image compression. The experimental
analysis is mainly performed on SAR data from various satellite projections. Thus, images were
processed in two different ways, analysis shows SBWC makes images smoother and conserves
object edges without affecting the resolution. In ending, the Segmentation based wavelet
compression technique allows output as expertly compressed image, with different operative
importance without changing the pixel accuracy and image resolution or size. The analysis
presented here will prove useful in studies of nonstationarity in time series, and the addition of
statistical significance tests will improve the quantitative nature of wavelet analysis.
7. REFERENCES
[1] Birgir Bjorn Saevarsson,Johannes R.Sveinsson and Jon Atli Benediktsson “Combined
Wavelet and Curvelet Denoising of SAR Images” Proceedings of IEEE 2004.
[2] A.Bruce and H, Gao. .Applied Wavelet Analysis with S-Plus.. Springer .Verlag New York, Inc.
1996.
[3] S. G. Chang, B Yu and M Vetterli. .Adaptive Wavelet Thresholding for image Denoising and
Compression.. IEEE Transactions on Image Processing, Vol. 9, No. 9, September 2000.
[4]. Gersho A., Gray M. “ Vector quantization and Signal compression”, Kluwer Academic
Publishers, Boston, 1992.
[5] M. Grgic, M. Ravnjak, and B. Zovko-Cihlar, “Filter comparison in wavelet transform of still
mages,” in Proc. IEEE Int. Symp. Industrial Electronics, ISIE’99, Bled, Slovenia, pp. 105–110.
1999,

[6] Guozhong Chen,Xingzhao Liu “Wavelet-Based Despeckling SAR Images Using Neighbouring
Wavelet Cofficients.” Proceedings of IEEE 2005.
[7] Guozhong Chen,Xingzhao Liu “An Improved Wavelet-based Method for SAR Images
Denoising Using Data Fusion Technique”. Proceedings of IEEE 2006.
[8]] J. Lu, V. R. Algazi, and R. R. Estes, “Comparative study of wavelet image coders,” Opt.
Eng., vol. 35, pp. 2605–2619, Sept. 1996.
[9] Mario Mastriani “New Wavelet-based Superresolution Algorithm for Speckle Reduction in
SAR Images” IJCS volume 1 number 4, 2006.
[10] Mulcahy, Colm. .Image compression using the Haar wavelet transform.. Spelman Science
and Math Journal. Found at: http://www.spelman.edu/~colm/wav.html.
[11] Sonka, M. Hiaual, V. Boyle, R. Image Processing, Analysis and Machine Vision, 2nd
edition. Brooks/Cole Publishing Company.
[12] William B., Joan L, “Still Image Data Compression Standard”, Van Nostrand Reinhold, New
York,1992
[13] www.mdpi.com/journal/sensors Article Haiyan Li 1,2 , Yijun He 1,* and Wenguang Wang
Improving Ship Detection with Polarimetric SAR based on Convolution between Co-
polarization Channels.
[14] Zhaohui Zeng and Ian Cumming,” SAR Image Compression Based on the Discrete Wavelet
Transform”, Presented at the Fourth International Conference on Signal Processing ICSP'98,
Beijing, China, October 12-16, 1998.
[15] T. Chang and C. Kuo, "Texture Analysis and Classification with Tree-Structured Wavelet
Transform", IEEE Trans. Image Processing, Vol. 2, No. 4, pp. 429-441, October 1993.
[16] Chang.S and B.Yu, “Spatially adaptive wavelet thresholding with context modeling for image
denoising.IEEE trans Image processing 9(9):533-539 ,2000.
[17] S. Arivazhagan and L. Ganesan. Texture Segmentation Using Wavelet Transform. Pattern
Recognition Letters, 24(16):3197–3203, December 2003.
[18] M. G. Mostafa, T. F. Gharib, and coll. Medical Image Segmentation Using a Wavelet-Based
Multiresolution EM Algorithm. IEEE International Conference on Industrial Electronics
Technology & Automation, December 2001.

Segmentation Based Multilevel Wide Band Compression for SAR Images Using Coiflet Wavelet

Recommended

Recommended

More Related Content

What's hot

What's hot (17)

Similar to Segmentation Based Multilevel Wide Band Compression for SAR Images Using Coiflet Wavelet

Similar to Segmentation Based Multilevel Wide Band Compression for SAR Images Using Coiflet Wavelet (20)

Recently uploaded

Recently uploaded (20)

Segmentation Based Multilevel Wide Band Compression for SAR Images Using Coiflet Wavelet